An Evaluation of the Reliability of State Personal Income Matthew von Kerczek Association for University Business and Economic Research Indianapolis, IN October 9, 2011
Overview of BEA estimates of State Personal Income ▪ BEA makes annual and quarterly estimates of state personal income. ▪ The annual estimates are published about 9 months after the end of the referenced year and are revised in each of the next two successive years. ▪ The annual estimates are more detailed than the quarterly estimates and are based on more complete source data. ▪ The quarterly estimates are published 3 months after the end of the referenced quarter and are revised three months later. ▪ The quarterly estimates are revised again when annual estimates are published. 2
Three Vintages of Estimates of State Quarterly Personal Income ▪ Preliminary Estimates: Current quarter wages based on change in Employment by NAICS sector from Current Employment Statistics (CES). Farm Proprietors’ Income NIPA controls ▪ Revised Estimates: Wages and Salaries from Quarterly Census of Employment and Wages (QCEW). Farm Proprietors’ Income Revised NIPA controls ▪ Latest Available Estimates: Revised NIPAs from Annual Comprehensive Revision Revised Annual State Estimates Revised Quarterly Source Data 3
How Reliable are BEA’s Preliminary Estimates of State Quarterly Personal Income ? ▪ For policymakers and businesses to better understand the economy and to formulate appropriate policies, timely and accurate estimates of State Personal Income are essential. ▪ The revised (second) and latest (third) estimates incorporate progressively more complete and more accurate source data as they become available from the Bureau of Labor Statistics, the U.S. Department of Agriculture and other sources. ▪ Successive estimates provide a consistent measure of economic activity over time and across states ▪ While there is always a trade-off between timeliness and accuracy, BEA believes that the preliminary State Quarterly Personal Income estimates have struck an appropriate balance, as the magnitude of the revisions to the preliminary estimates tends to be small. 4
Reliability Criteria ▪ Reliability refers to the ability of successive vintages of estimates to present a consistent picture of economic activity within a state. ▪ Mean Revision (MR) = sum of (L-E)/n, where: L is the rate of change in the ‘Latest Available’ estimates E is the rate of change in the earlier vintage n is the number of observations in the sample period over which the mean is calculated. ▪ Mean Absolute Revision (MAR) = sum of |L-E|/n: Also can be thought of as the mean revision without regard to sign. 5
Periods / Scope of Study ▪ Last two complete business cycles (expansion to trough). 1991:II-2001:IV 2002:I-2009:II ▪ Scope of Study: Total Personal Income Nonfarm Personal Income Total Wages and Salaries ▪ The analysis that follows is based on BEA estimates of state personal income as of June 21,
Results (Revisions to rates of change) 7 Mean Absolute Revisions, Latest Estimates less Preliminary and Second Estimates 2002:I-2009:II (ranked by MAR to Preliminary Estimates of Personal Income) Personal IncomeNonfarm Personal IncomeTotal Wages and Salaries PreliminarySecondPreliminarySecondPreliminarySecond United States Mideast New England Southeast Great Lakes Plains Southwest Rocky Mountain Far West Mean Absolute Revisions, Latest Estimates less Preliminary and Second Estimates, 1991:II-2001:IV (Ranked by MAR to Preliminary Estimates of Personal Income) Personal IncomeNonfarm Personal IncomeTotal Wages and Salaries PreliminarySecondPreliminarySecondPreliminarySecond United States Plains Rocky Mountain New England Great Lakes Southwest Far West Mideast Southeast
Results (Revisions to rates of change) Mean Absolute Revisions, Latest Estimates less Preliminary estimates (Ranked by MAR to Preliminary Estimates of Personal Income) 2002:I-2009:II Personal Income Nonfarm PIWages Largest revisions Louisiana North Dakota South Dakota Wyoming Iowa Nevada New York Nebraska Mississippi United States Smallest revisions Virginia Ohio West Virginia Oregon Maryland Pennsylvania Kentucky Alabama South Carolina Mean Absolute Revisions, Latest Estimates less Second estimates (Ranked by MAR to Second Estimates of Personal Income) 2002:I-2009:II Personal IncomeNonfarm PIWages Largest Revisions Louisiana North Dakota South Dakota Mississippi Nebraska Iowa Wyoming United States Smallest Revisions Illinois South Carolina Ohio Pennsylvania Maryland California Tennessee
Aggregate Average Revisions (to rates of change) by Quarter Mean Absolute Revisions, Latest Estimates less Preliminary and Second Estimates, 2002:I-2009:II [Percentage Points] Personal IncomeNonfarm Personal IncomeTotal Wages YRQTRPreliminarySecondPreliminarySecondPreliminarySecond
Range of Revisions (to rates of change) 10
Measuring Direction of Change & Acceleration / Deceleration Reliability of Quarterly Estimates of State Personal Income, 2002:I :II Percent with Correct Indication Direction of Change Acceleration or Deceleration From the Previous Quarter PreliminarySecondPreliminarySecond United States97% 63%77% New England90% 67%73% Connecticut83%80%63%57% Maine93% 70%83% Massachusetts87% 70%83% New Hampshire83% 70%80% Rhode Island83%93%73%87% Vermont93% 77%83% Mideast93%90%73%87% Delaware90%87%73%70% District of Columbia93% 87% Maryland93%97%87%93% New Jersey93% 63% New York83%80%70%77% Pennsylvania90% 70%80% Great Lakes93%90%77%83% Illinois93%97%70%73% Indiana93% 77% Michigan87%97%80%77% Ohio93% 73%90% Wisconsin93% 53%70% Plains 93% 57%70% Iowa87% 47%60% Kansas93%90%67%63% Minnesota97%90%77%87% Missouri93%97%67%73% Nebraska83% 60%57% North Dakota83% 63%70% South Dakota80%83%60% Reliability of Quarterly Estimates of State Personal Income, 2002:I :II Percent with Correct Indication Direction of Change Acceleration or Deceleration From the Previous Quarter PreliminarySecondPreliminarySecond Southeast97% 60%73% Alabama97%100%80%87% Arkansas93%90%53%43% Florida97%100%83%90% Georgia97%100%67%70% Kentucky97% 67%77% Louisiana90% 50%67% Mississippi87%90%67%73% North Carolina97% 63% South Carolina97%100%73%83% Tennessee97%100%60%63% Virginia93%100%70%73% West Virginia93%97%50%57% Southwest97%93%47%60% Arizona97%100%93%97% New Mexico90%93%63%70% Oklahoma87%93%53%63% Texas97%90%53%57% Rocky Mountain97% 70%83% Colorado93%87%83%80% Idaho93%90%70%83% Montana93%97%57%67% Utah93%100%73%77% Wyoming90%93%63%77% Far West97%100%70%90% Alaska97% 77% California97% 67%87% Hawaii93%100%70% Nevada87%93%77%90% Oregon100% 70%77% Washington St97% 67%77% 11
Recent Developments ▪ In 2002 BEA began incorporating QCEW wages into the quarterly NIPA estimates of wages and salaries (private) on a concurrent basis. ▪ In 2004 BEA accelerated the release of State Quarterly Personal Income. ▪ In 2009 BEA substantially redefined the treatment of uninsured property losses in the NIPAs. ▪ Improvements in Farm Income 12
Preliminary Observations ▪ Overall, the preliminary and second estimates of quarterly estimates of State Personal Income are reliable in predicting direction of change (97 percent of the time) and whether the change is accelerating or decelerating (about 70 percent of the time). ▪ Overall, the revisions to the preliminary estimates of personal income, nonfarm personal income, and total wages were slightly higher in the period. ▪ However, the revisions to the second estimates of personal income, nonfarm personal income, and total wages were considerably lower in the period. ▪ Revisions to Farm Income continue to cause outsized revisions to states where farming is important, but even among farm states revisions were lower in the second period. Revisions to farm income were significantly smaller in the later time period. ▪ Aside from the revisions in states where farming is important, most of the revisions to personal income were driven by revisions to wages and salaries. ▪ Unpredictability of lump-sum payments (bonus, stock etc.) affecting reliability of preliminary wage estimates. ▪ Conceptual definition changes to the treatment of uninsured losses from natural disasters affected the reliability measures in Louisiana and Mississippi. 13
Contact information Matthew A. von Kerczek Economist Regional Income Division (202)